AMICA1 is down-regulated in LUAD and may function as a diagnostic biomarker
We first assessed AMICA1 mRNA expression level in GSE116959 (T = 57, N = 11, P < 0.001, Fig. 1A), GSE32867 (T = 145, N = 144, P < 0.001, Fig. 1B), GSE43458 (T = 80, N = 30, P < 0.001, Fig. 1C) and TCGA database (T = 437, N = 54, P < 0.001, Fig. 1D), and the results obtained showed that, compared with normal tissues, LUAD patients’ AMICA1 expression was obviously lower. We further investigated the AMICA1 expression level in LUAD tissues and matched paracancerous non-cancerous tissues in GSE32863 (T = 60, N = 60, P < 0.001, Additional file 1: Fig. S1A), GSE75037 (T = 83, N = 83, P < 0.001, Additional file 1: Fig. S1B) and TCGA (T = 49, N = 49, P < 0.001, Additional file 1: Fig. S1C) database through Wilcoxon single rank test, and discovered that the AMICA1 expression was also significantly decreased in LUAD tissues. These findings implied that AMICA1 may play an inhibitory role in the LUAD development. Then, we further investigated the value of AMICA1 in LUAD diagnosis. The ROC curve of GSE116959 (Fig. 1E), GSE32867 (Fig. 1F), GSE43458 (Fig. 1G) and TCGA database (Fig. 1H) showed that AMICA1 can function as a potential diagnostic marker for LUAD, the AUC values were 0.927, 0.982, 0.797 and 0.956, respectively. And the sensitivity was 0.825, 0.948, 0.867, 0.932 and the specificity was 1.000, 0.914, 0.713 and 0.907, respectively. As for the paired LUAD and paracancerous non-cancerous tissues, that is GSE32863 (Additional file 1: Fig. S1D), GSE75037 (Additional file 1: Fig. S1E) and TCGA database (Additional file 1: Fig. S1F), we also calculated the AUC values, they were 0.982, 0.991 and 0.952, respectively. The sensitivity was 0.948, 0.952, 0.932 and the specificity was 0.914, 0.964 and 0.881, respectively. All of these results showed that AMICA1 was down-regulated in LUAD tissue and may function as a biomarker for LUAD diagnosis.
Correlations between AMICA1 expression and clinicopathological parameters in LUAD patients
Since AMICA1’s function in LUAD is still unclear, it is necessary to further explore the connections between the expression level of AMICA1 and the clinical parameters in LUAD patients. Thus, the Wilcoxon test and Kruskal test were applied to analyze the correlation between AMICA1 expression level and different clinicopathological features in LUAD patients. TNM stage is the most widely used method for tumor stage. T, N, M represents the status of primary tumor, lymph-node metastasis and distant metastasis, respectively. The detailed TNM stage can be seen in the eighth edition IASLC lung cancer stage project [11,12,13]. TCGA data showed that AMICA1 expression was linked to T stage (P < 0.001, Fig. 2B), M stage (P = 0.046, Fig. 2D) and TNM stage (P = 0.007, Fig. 2A) of LUAD. These results revealed that AMICA1 was significantly decreased in advanced LUAD patients.
Furthermore, we used Cox regression to analyze the prognostic role of AMICA1 in LUAD. The univariate analysis showed that low AMICA1 expression was associated with worse overall survival (OS) (P = 0.006, Additional file 2: Fig. S2A). Besides, as expected, the clinical parameters, such as advanced T stage (P < 0.001), N stage (P < 0.001), M stage (P = 0.019) and TNM stage (P < 0.001), were all related to worse OS (Additional file 2: Fig. S2A). To further verify AMICA1’s prognostic value in LUAD, multivariate analysis was performed. The result obtained revealed that only AMICA1 expression (P = 0.019) and TNM stage (P = 0.006) were independently related to OS (Additional file 2: Fig. S2B), which means that the role of AMICA1 in evaluating patients' clinical prognosis is superior to T stage, N stage and M stage.
AMICA1 is related to immune infiltration level and LUAD prognosis
Next, we explored whether the expression level of AMICA1 was linked to various immune cell infiltration in LUAD from TIMER database. Pearson correlation analysis displayed a significantly positive connection between AMICA1 expression and B cell (R = 0.48, P < 0.001), CD4+ T cells (R = 0.54, P < 0.001), CD8+ T cells (R = 0.42, P < 0.001), dendritic cells (R = 0.69, P < 0.001), macrophages (R = 0.43, P < 0.001) and neutrophil (R = 0.57, P < 0.001) (Fig. 3A). The positive correlations between AMICA1 expression and these immune cells in the TCGA-LUAD dataset were also well confirmed in GSE72094 dataset (Fig. 3B).
We then used the ESTIMATH algorithm to analyze whether AMICA1 expression was associated with the total level of immune cells infiltration in LUAD. The results obtained showed a positive connection between the expression level of AMICA1 and immune score in both TCGA (P < 0.001, Fig. 3C) and GEO LUAD datasets (P < 0.001, Fig. 3D). Moreover, LUAD patients with high immune scores have a better OS (Fig. 3E and F), which was consistent with the prognostic results of AMICA1 (Fig. 3G and H).
To further broaden the cognition of the correlation between AMICA1 and immune infiltration, we analyzed the connections between AMICA1 and tumor-infiltrating lymphocytes (TILs), immunomodulators, chemokines and related receptors, and respectively listed the top six related cells and molecules. Spearman associations test between AMICA1 expression and various immune signatures were obtained from the TISIDB database. Additional file 3: Fig. S3A (left) shows the correlations between AMICA1 expression and 28 TILs abundance in 30 kinds of cancers. The results obtained showed that AMICA1 was related to many TILs in LUAD, the top six TILs (Additional file 3: Fig. S3A right) were respectively macrophage (r = 0.754, P < 0.001), CD8+ effector memory T cells (TEM_CD8) (r = 0.744, P < 0.001), myeloid-derived suppressor cells (MDSCs) (r = 0.735, P < 0.001), mast cells (r = 0.734, P < 0.001), follicular helper T cells (Tfhs) (r = 0.726, P < 0.001) and immature B cells (Imm_B cells) (r = 0.691, P < 0.001). Additional file 3: Fig. S3B–F shows the correlations between AMICA1 and 24 immunoinhibitors, 45 immunostimulators, 21 MHC molecules, 41 chemokines and 18 related receptors, respectively.
Finally, we analyzed the expression of AMICA1 in infiltrating immune cells subtype of LUAD tissues with single-cell sequence data from GSE131907 (neutrophils were not recovered in the experimental process because of technical reasons). All of the patients were treatment-naïve and were divided into early- and advanced- stages. The early-stage was defined as the patients without lymph nodes and distant metastasis. We found out that AMICA1 was mainly expressed in myeloid cells in early- and advanced- stage LUAD tissues (Early-stage LUAD: Additional file 4: Fig. S4A, B, Advanced-stage LUAD: Additional file 4: Fig. S4E, F). Then, we further analyzed the expression of AMICA1 in different myeloid cells subtypes, including monocytes, macrophages (mo-Mac, Alveolar Mac and Pleural Mac) and dendritic cells (CD1c+ DCs, CD207+ CD1a+ LCs, CD163+ CD14+ DCs, Activated DCs, CD141+ DCs and pDCs). Interestingly, the expression of AMICA1 is significantly decreased in mo-Mac compared to other myeloid cell subtypes, including monocytes, alveolar Mac and aleural Mac and is significantly increased in CD1+ DCs compared to other DCs subtypes (Early-stage LUAD: Additional file 4: Fig. S4C, D, Advanced-stage LUAD: Additional file 4: Fig. S4G, H). Different from alveolar Mac and aleural Mac, Mo-Mac are monocyte-derived macrophages and could create an immunosuppressive microenvironment [11], which means that decreased AMICA1 expression might play an important role in the formation of mo-Mac and immunosuppression.
AMICA1 co-expression networks in LUAD
To further explore AMICA1’s biological significance in LUAD, AMICA1 co-expression network was investigated through the function module of LinkedOmics. Volcano plots in Fig. 4A show that 3207 genes (dark red dots) were positively related to AMICA1, and 1292 genes (dark green dots) were negatively related. Figures 4B and C show respectively the top 50 significant genes that are positively and negatively related to AMICA1. Notably, in the top 50 significantly positive genes, there were 36 genes with a low hazard ratio (HR) (P < 0.05), which means that they may function as the low-risk genes like AMICA1. In contrast, there were 24/50 genes with high a HR (P < 0.05) in the top 50 negatively significant genes (Fig. 4D).
Besides, GO-BP term annotation by GSEA demonstrated that AMICA1 co-expressed genes join mainly in respiratory burst, interleukin-1 production, interferon-gamma production, cellular defense response, leukocyte activation involved in inflammatory response, adaptive immune response, interleukin-4 production, leukocyte proliferation, tumor necrosis factor superfamily cytokine production and neuroinflammatory response, etc. (Fig. 4E). And KEGG pathway analysis illustrated enrichment in autoimmune thyroid disease, asthma, allograft rejection, staphylococcus aureus infection, hematopoietic cell lineage, graft-versus-host disease, intestinal immune network for IgA production, leishmaniasis, type I diabetes mellitus and viral myocarditis, etc. (Fig. 4F).
These results show a wide influence of AMICA1 on the immune response and prognosis in LUAD patients.
Validation of AMICA1 expression and the correlation with immune infiltration
To further confirm the above results, we analyzed AMICA1 mRNA and protein expression levels in LUAD tissues and adjacent non-tumor tissues and several lung cancer cell lines (A549, H1437 and H460) by qRT-PCR and western blotting. Besides, the CD8+ T cells and PD1+ T cells infiltration levels were detected via immunohistochemistry (IHC). The results obtained show that the mRNA and protein level of AMICA1 were both significantly decreased in LUAD tissues when compared with adjacent non-tumor tissues (Fig. 5A and D). And the ROC curves revealed that AMICA1 can function as a diagnostic marker with AUC = 0.803, Sensitivity = 0.538 and Specificity = 1 (Fig. 5B). Besides, the mRNA and protein level of AMICA1 were both significantly decreased in A549, H1437 and H460 lung cancer cell lines when compared with human normal pulmonary epithelial cell Beas2B (Fig. 5C and E). The IHC results and spearman correlation test confirmed that the expression level of AMICA1 was positively correlated with the infiltration levels of CD8+ T cells (P = 0.002) and PD1+ T cells (P = 0.003, Fig. 5F, G).
Overexpression of AMICA1 suppressed the proliferation of LUAD and activated cGAS-STING signaling
To determine the biological role of AMICA1 in LUAD proliferation, we overexpressed AMICA1 in A549 and H1437 cell line, the overexpression efficiency was confirmed by qRT-PCR and western blot (Fig. 6A, B). The results of CCK8 (Fig. 6C, D) and EDU (Fig. 6E, H) suggested that overexpression of AMICA1 significantly suppressed the proliferation ability of A549 and H1437 cells. Besides, given that cyclic GMP-AMP synthase and stimulator of interferon genes (cGAS-STING) signaling is important to immune infiltration of LUAD, we investigated the correlation between AMICA1 and cGAS-STING signaling from TCGA database, including MB21D1, TMEM173, IRF3 and TBK1. The results obtained show that the mRNA expression level of AMICA1 in LUAD tissues was positively correlated with TMEM173 (P < 0.001, cor = 0.49, Fig. 7A) and IRF3 (P = 0.027, cor = − 0.092, Fig. 7A). Our results showed that overexpression of AMICA1 could increase the expression of TMEM173 mRNA level (Fig. 7B, C) and protein level (Fig. 7D and F), but had no influence on the expression of MB21D1, IRF3 and TBK1. The result of immunofluorescence also showed that TMEM173 was significantly up-regulated after the overexpression of AMICA1 (Fig. 7H, I). Besides, we further analyzed the nuclear/cytoplasmic fluorescence ratio and found that AMICA1 had no influence in the location of TMEM173 (Fig. 7J). Given that the phosphorylation of TMEM173 is necessary for the activation of cGAS-STING signaling, we then detected the phosphorylation level of cGAS-STING signaling. The results obtained show that the phosphorylation levels of TMEM173, IRF3 and TBK1 were up-regulated in A549 and H1437 cell lines (Fig. 7E and G). Besides, the expression of type I interferon was up-regulated as well (Fig. 7E and G).